Memory is Cue Driven: Functional Vs Nominal cues

‘whilst part of what we perceive comes through our senses from the object before us, another part always comes out of our own mind’ – William James (1890, vol.2, pp. 103)

Tulving (1983) suggested that memory traces (‘engrams’) possess functional

properties, rather than structural ones, and they should be described in terms of what they do, and not what they are. According to this, the material presented to the participant is known as nominal stimuli (cues or targets). What an experimenter considers as an important retrieval cue can only be thought of as a nominal retrieval cue. The component of the nominal stimulus that the participant actually encodes (and any further processing of that cue that involves knowledge or other stored information) becomes the effective cue for response elicitation, and constitutes the functional cue (Capaldi & Neath, 1995; Postman, Stark &

Fraser, 1968; Underwood, 1963). The physical characteristics of an object may not change, but what is perceived differs depending on the observer’s point of view. What is ultimately stored in memory does not only depend on the material and its characteristics, but also on the characteristics of the context and on the identification and processing of the stimuli. One’s cognitive state provides the context within which the stimulus is interpreted (Bower, 1972b).

A retrieval cue appears to be the product of a person’s mental activity during encoding, but also during retrieval; depending on these activities, the same stimulus may have different cueing functions (e.g. Mathews, 1977). Tulving and Thomson (1971) referred to all the factors present at the time of learning, other than the event, that influence encoding as

‘cognitive environment’. Processing the available information within a specific cognitive

environment determines the nature of the functional cue, either by adding, or subtracting (or both) from the nominal cue (Neath & Surprenant, 2003).

Cue Processing

The specific processing of the information will determine the nature of the cues, but experimental manipulations can only partially affect the encoding processes, and they can never completely control them. Postman, Adams and Phillips (1955) used an early version of the encoding/retrieval paradigm to investigate encoding processes via intentional and

incidental learning (a later version was used by Tulving and colleagues to explore the interactions between encoding and retrieval processes e.g. Tulving and Osler, 1968; Tulving and Thomson, 1973). Participants were shown a list of 30 adjectives and were asked to judge the frequency of occurrence of each one. In the intentional condition, they were also told their memory for the adjectives would later be tested, while in the incidental condition they were told nothing. Both groups (intentional and incidental) were divided into three testing

conditions that manipulated the task difficulty, as well as the context at the time of retrieval (facilitating versus inhibitory). The first condition employed a free recall task (difficult condition), while the other two a cued-recall task (easy conditions). The manipulation of the retrieval context occurred in the cued-recall conditions; in one testing condition, the cues were extra-list words that were closely associated with the target adjective, and therefore would facilitate responses (facilitating retrieval context). In the other testing condition, the cues were remotely associated with the targets and were expected to elicit competing

responses (inhibitory retrieval context). Incidental learning led to lower memory performance only in the free recall test condition. Performance of intentional and incidental learners did not differ in, either the facilitating, or the inhibitory cued-recall condition. The presence of semantically related extra-list cues (facilitating retrieval context) boosted performance, but the presence of unrelated words harmed retrieval relatively to free recall. The authors

concluded that the mental activity during the learning affects subsequent ease of recall of the material irrespective of the instructions. Postman (1964) examined the literature on incidental learning and suggested that intention is not the determining factor of performance. Rather, some unobservable internal processes – triggered by the learning instructions and correlated with the following mental activity - establish the extent to which the material will be

subsequently remembered. It could be the case that pre-existing associations may be responsible for the obtained results.

Based on an associative view of memory (see Anderson and Bower, 1974 for a

detailed account) it is expected that a cue, A, that has acquired an association with a response, B, will become an effective retrieval cue. The association is established due to simultaneous occurrence, and its strength is determined by frequency, recency, meaningfulness, etc.

(Tulving, 1983). Tulving and Pearlstone (1966) found cued-recall to be superior to free recall. The effectiveness of category names in improving memory for a list of words can be explained by the pre-existing associations between the target words and the category, which was further strengthened by their co-occurrence in the list. This explanation though fails to account for other results, such as those obtained by Underwood, Runquist and Schulz (1959).

They also compared free recall of response members of a associate list with paired-associate (cued) recall. This comparison was similar to the one made by Tulving and Pearlstone (1966) with one difference: the type of materials used. Underwood et al. (1959) paired nonsense syllables (cues) with adjectives (targets), thus precluding the existence of any pre-experimental associations. Unlike the previous superiority of cued recall, this time it was free recall that led to higher performance. This discrepancy between two essentially similar experiments, questioned the conditions under which a retrieval cue becomes adequately effective to increase memory performance relatively to a free recall task. In a series of experiments, Tulving and colleagues tried to determine the extent to which

pre-experimental associations are responsible for the superiority of cued-recall over free recall (for a detailed account see Tulving, 1983). They concluded that those associations could not always predict a better performance during cued-recall compared to free recall. It appears that the pre-experimental associations and their strength are not sufficient to explain memory performance, and some other factor(s) comes into play.

Most experiments investigating the effect of encoding operations on cued-recall did not control for pre-existing associations between the cues and the target (e.g. Fisher and Craik, 1977; Moscovitch and Craik, 1976), which made it impossible to preclude it as a possible explanation. Mathews (1977) used word triplets to explore encoding operations in cued recall. He attempted to hold the nominal identity of the cue and target words, their pre-existing associations, and their co-occurrence during learning constant in all conditions.

Thus, any observed effects could be attributed to the different mental processes occurring during the encoding of each condition. Each triplet contained two nouns that were

semantically related, and members of a broader conceptual category (e.g. aluminium, cement). The third word was the name of a category and there were three encoding manipulations. In the similarity condition, the third word would indicate the conceptual category of which both noun words were members (e.g. building material). In the contrast condition, the third word would represent a category of which just one of the noun words was member (e.g. metal). In the negative encoding condition, none of the noun words were

members of the category designated by the third word (e.g. beverage). During learning, the participants had to answer whether both, one or none of the nouns were related to the category. Each noun pair was presented once with each of the three category words. Half of the participants were informed before making these judgments about a subsequent memory test, in which they would be given any one of the three words and they would have to retrieve the other two. Performance did not differ between incidental and intentional learning

instructions, but recall differed as a function of the encoding operations: a noun cue was most effective in retrieving the other noun under the similarity encoding condition, while it was moderately effective under the contrast condition, and not effective at all under the negative condition. This result stresses the importance of the encoding operations during presentation in determining the effectiveness of the same nouns as cues: what was compared across conditions was how well one noun cued the other but the identity of the nouns and pairs did not change across conditions. The only things that changed, in theory, were the identity of the category and the encoding operations that the instructions generated. It appears that a

retrieval cue is the product of a person’s mental activity during encoding, but also during retrieval and that depending on these activities the same stimulus may have different cueing functions (e.g. Mathews, 1977). ‘The effectiveness of retrieval cues depends directly on how the to-be-remembered word is stored, and only indirectly on the pre-experimentally

established relations between the cue and the to-be-remembered target’ (Tulving, 1983, pp.

210).

Cue Selection

The processing during encoding has an impact on the effect of the individual cues during retrieval. The previous section examined cases where the type of processing was manipulated by the experimenter (incidental learning or not, focused attention to certain material, different encoding instructions). But people may also select to process some

material more (or less) at their will creating effective functional cues. There was considerable interest in exploring this pre 1980s, while, to our knowledge, there was not much research relating specifically to theories of cue selection after that. This section briefly reviews the relevant literature on what could possibly guide cue selection before moving on to what may determine cue effectiveness. This chapter concludes with a suggestion of what may be a significant guiding force of both.

Underwood, Ham and Ekstrand (1962) suggested that participants in a memory task, when faced with a complex or compound stimulus, would select one of its components / features to use as an effective cue (functional cue) for the retrieval (response). According to Underwood et al., learning in a paired-associate task, where the cue has two components, can occur in different ways depending on cue selection processes. They assumed that, because of cue selection, there would be a discrepancy between the nominal and the functional cue.

Some support for this claim had been previously provided by Underwood & Schulz (1960), when considering participants’ verbal reports. After a paired-associate task, the participants described how they used part of the original three-letter stimulus (e.g. a single letter instead of all three letters) as a cue for the target.

If cue selection occurs, and just one part of a stimulus becomes the functional cue, what are the stimulus features that are selected in order to construct that cue?

One attribute that has been speculated to affect cue selection is meaningfulness (Solso, 1968; 1971). Sundland and Wickens (1962) investigated the learning of paired associates in a particular context (background colour), and the subsequent target recall performance with or without the context. The cues were either high frequency words, or nonsense syllables, and the targets were always words. After every five cue-target

presentations each in a different background colour, anticipation trials followed, where the cue, the colour or both were presented, and participants were instructed to try to predict the target. The number of correct responses at the anticipation trials was used as indication of learning performance. Two groups were tested under each of the following learning conditions: colour (c), word (w), nonsense syllable (ns), word-colour (w+c) and nonsense syllable-colour (ns+c). After the 15^th cue presentation and the regular anticipation trial, the word-colour and the nonsense syllable-colour groups received two additional trials. One group from each learning condition was tested with the colour list first, and then with the

symbol (word or syllable) list, while the order was reversed for the other group. Recall performance was expressed as percentage: number of correct recalls divided by the correct responses in the last anticipation trial.

The context (background colour) was expected to facilitate learning of the less meaningful cues (nonsense syllables), but not to have an effect on learning of the highly meaningful cues (words). Moreover, recall using the associated context was expected to be better for nonsense syllables, than words. Results showed no effect of the different coloured background on the learning of either the words, or the nonsense syllables. Nevertheless, the associations between the coloured background and the cues were better for the nonsense syllables, than for the words: at the additional trials, removing the context (background colour) did not affect recall performance for the word group (97.8%), but harmed recall for the nonsense syllable group (72.8%). On the contrary, when the symbol was removed (word or nonsense syllable), performance severely deteriorated for the word group (10.8%), but not so dramatically for the nonsense syllable group (47.8%).

In a second experiment, participants were also asked to describe the stimuli they used during learning. Verbal reports indicated that very few participants used both the symbol and colour dimension of the cue. The majority used one or the other alone, or alternated between them, as if just one dimension was available. Most participants in the colour-word group used the symbols (words), while in the colour-nonsense syllable group participants used the colour alone. It seems that the highly meaningful symbols (words) overshadowed the use of the context as a potentially effective cue. In the case of less meaningful symbols though, the context resumes its value as a response cue to such an extent that it is preferred over the symbol.

Underwood, Ham and Ekstrand (1962) explored further whether the more meaningful component of a compound stimulus would become the functional cue. They presented two

lists of compound cues, consisting of a verbal unit and a colour, paired with numbers as targets. In one of the lists, the verbal units were low-meaning trigrams, while in the other they were common three-letter words. Following learning of the lists (one perfect recitation), a second paired-associate list was given to the participants, including just part of each compound cue along with the targets (transfer list). Overall, there were four groups with a different combination of initial and transfer paired-associate lists (initial list + transfer list):

word-colour + colour, word-colour + word, trigram-colour + colour and trigram-colour + trigram. In addition, there were two control conditions: word-colour + word-colour and trigram-colour + trigram-colour. It was expected that, if the more meaningful component of a nominal compound cue becomes the functional cue, recall performance in the transfer list would be reduced in the conditions were that component was removed. Assuming that colours are more meaningful than trigrams, it was hypothesised that the colour would be selected as the functional cue in the first list. However, in the second list it was expected that words would be selected over colours. It was therefore predicted, that if the trigrams were removed from the trigram-colour compound, there would be little effect on paired-associate performance in the transfer list. The opposite was predicted if the colour was removed from the same compound. Inversely, for the word-colour compounds, performance would be greatly affected in the transfer list by the removal of the words, but not from the removal of the colours. Results showed that, in the trigram-colour list, colour was indeed the most effective cue in eliciting correct target responses in the transfer list with the removal of trigrams having only a small effect. For the word-colour list, the presence of the words alone in the transfer lists induced more correct target responses, than the presence of the colours alone. The latter result could be due to participants’ preference, in general, towards verbal material (as opposed to the colour patches used), and not entirely due to their increased

meaningfulness. But the preference of colours to trigrams as functional cues was attributed to their higher meaningfulness.

Spear, Ekstrand, & Underwood (1964) replicated the above finding in a paired-associate task using compound stimuli, consisting of a trigram and a word, as cues and

numbers as targets. Group W was only shown the words as cues in the transfer list, and group T was only shown the trigrams. It was expected that the more meaningful element of the compound (i.e. the word) would elicit the highest performance. Results showed that group W maintained a high level of performance throughout the transfer tests (mean of 75.75 total correct responses). Group T showed a negatively accelerated learning curve but was overall significantly below group W (mean of 52.08 total correct responses). This finding suggests that cue selection occurred, and that the most meaningful element of the compound stimulus, the words, was chosen. It seems that the more meaningful a feature is, the more probable it is that it will be selected as a functional cue.

Another attribute thought to be affecting cue selection is the degree of formal

similarity among the cue features. Cohen and Musgrave (1966) created six lists of cue-target pairs with compound cues and single letters as targets. Each compound cue consisted of two nonsense syllables. In a third of the lists (mixed compound lists), one of the syllables had high formal similarity (H) with other syllables within the cue set e.g. ryg, byg, gyr, gyb, and the second syllable had low formal similarity (L) relatively to the other cue syllables e.g. tep, muz, cah, zil. In another third, the two syllables were both of high similarity (HH) and the last set of lists consisted of low similarity syllables (LL). They hypothesised that, in the mixed compound cue list (HL), the high similarity syllable would be ignored, and

participants would select the relatively distinctive syllable, and associate it with the response.

Results on ease of learning showed that it was easier to learn the low similarity compound lists (LL), followed by the mixed compound list (HL), and then the purely high similarity

compound list (HH). In the transfer task, where only one element of the compound was shown, participants gave more correct responses to the low similarity syllables. Position of the elements also had an effect with those in the first position leading to higher accuracy. A similar result was found by Cohen & Musgrave (1964): low meaningfulness CVCs in the first position of compound cues were better learned, than when they were in the second position.

Participants tend to learn initially the elements in the first position, and later on those in the second. If the second position elements are more discriminable, then they are better learned, than the ones in the first position.

Rabinowitz and Witte (1967) explored whether a single distinctive element of a multiple element stimulus would be selected to become the effective cue for response elicitation. They created a paired-associate list of non-overlapping consonant-trigram cues with numbers (ranging from one to seven) as targets. One of the trigram letters was red, while the other two were black. The position of the red letter (first, second, third) was manipulated between participants. The paired associate task was used to test learning. The learning

criterion was two perfect repetitions after which a surprise transfer task followed. Participants were presented with each individual trigram letter, and were asked to assign the correct target number. Results did not show any difference in the learning rate of the lists. In the transfer task, there was a significantly higher performance (more correct responses) when the red letter was presented. It seems that participants use a single distinctive element (the red letter) of a multiple element nominal stimulus as the effective cue for responding. Examination of the letter position revealed that the first group (red letter in the first position) made

significantly more correct responses to the red letter, than to the black letters, but there was no difference between the second back and the third black letter. The second group (red letter

significantly more correct responses to the red letter, than to the black letters, but there was no difference between the second back and the third black letter. The second group (red letter